Extracting Lungs from CT Images via Deep Convolutional Neural Network Based Segmentation and Two-Pass Contour Refinement
- PDF / 5,665,031 Bytes
- 14 Pages / 595.224 x 790.955 pts Page_size
- 45 Downloads / 152 Views
ORIGINAL PAPER
Extracting Lungs from CT Images via Deep Convolutional Neural Network Based Segmentation and Two-Pass Contour Refinement Caixia Liu1 · Mingyong Pang1 Received: 13 December 2019 / Revised: 17 August 2020 / Accepted: 14 September 2020 © Society for Imaging Informatics in Medicine 2020
Abstract Lung segmentation is a key step of thoracic computed tomography (CT) image processing, and it plays an important role in computer-aided pulmonary disease diagnostics. However, the presence of image noises, pathologies, vessels, individual anatomical varieties, and so on makes lung segmentation a complex task. In this paper, we present a fully automatic algorithm for segmenting lungs from thoracic CT images accurately. An input image is first spilt into a set of non-overlapping fixed-sized image patches, and a deep convolutional neural network model is constructed to extract initial lung regions by classifying image patches. Superpixel segmentation is then performed on the preprocessed thoracic CT image, and the lung contours are locally refined according to corresponding superpixel contours with our adjacent point statistics method. Segmented lung contours are further globally refined by an edge direction tracing technique for the inclusion of juxta-pleural lesions. Our algorithm is tested on a group of thoracic CT scans with interstitial lung diseases. Experiments show that our algorithm creates an average Dice similarity coefficient of 97.95% and Jaccard’s similarity index of 94.48%, with 2.8% average over-segmentation rate and 3.3% under-segmentation rate compared with manually segmented results. Meanwhile, it shows better performance compared with several feature-based machine learning methods and current methods on lung segmentation. Keywords Lung segmentation · Deep convolutional neural network · Superpixel segmentation · Contour correction
Introduction Lungs, principal components of the respiratory system in human body, are highly susceptible to diseases, where lung cancer is one of the malignant tumors with highest morbidity and mortality [1]. Lung segmentation of CT images is a precursor to most pulmonary image analysis applications and it plays an important role in computeraided pulmonary disease diagnostics. However, accurate lung segmentation is still a challenging issue in thoracic CT image analysis due to lung shape variances, image noises,
Mingyong Pang
[email protected] Caixia Liu [email protected] 1
Institute of EduInfo Science and Engineering, Nanjing Normal University, No. 122, Ninghai Ave., Nanjing, 210097, People’s Republic of China
highly varied properties of pulmonary diseases, and so on (see Fig. 1). Research showed that 5–17% of lung nodules in their sampled data were missed due to the inaccurate lung segmentation [2]. Conventional methods for lung segmentation, such as thresholding, region growing, watershed, and active contour, depend largely on a large difference of grayscale values between lung regions and their surrounding tissues in thoracic CT images. The methods are easily in
Data Loading...
